Applied Soft Computing 4r3r3u

EE6006 - APPLIED SOFT COMPUTING

UNIT I: ARCHITECTURES – ANN

Introduction – Biological neuron – Artificial neuron – Neuron model – Supervised and unsupervised learning- Single layer – Multi layer feed forward network – Learning algorithm- Back propagation network. UNIT II : NEURAL NETWORKS FOR CONTROL networks – Discrete time Hopfield networks – Transient response of continuous time system – Applications of artificial neural network - Process identification – Neuro controller for inverted pendulum. UNIT III : FUZZY SYSTEMS Classical sets – Fuzzy sets – Fuzzy relations – Fuzzification – Defuzzification – Fuzzy rules hip function – Knowledge base – Decision-making logic – Introduction to neuro fuzzy systemAdaptive fuzzy system. UNIT IV : APPLICATION OF FUZZY LOGIC SYSTEMS Fuzzy logic control: Home heating system - liquid level control - aircraft landing- inverted pendulum – fuzzy PID control, Fuzzy based motor control.

UNIT V : GENETIC ALGORITHMS Introduction-Gradient Search – Non-gradient search – Genetic Algorithms: binary and real representation schemes, selection methods, crossover and mutation operators for binary and real coding - constraint handling methods – applications to economic dispatch and unit commitment problems.

BOOKS  Laurance

Fausett,

‘Fundamentals

of

Englewood Neural

cliffs,

N.J.,

Networks’,Pearson

Education,1992.  Timothy J. Ross, ‘Fuzzy Logic with Engineering Applications’, Tata McGraw Hill, 1997.

 S.N.Sivanandam and S.N.Deepa, Principles of Soft computing, Wiley India Edition, 2nd Edition, 2013.

INTRODUCTION VS

SOFT COMPUTING  Idea - To model cognitive behavior of human mind.  Conceptual intelligence in machines

 Tolerant of imprecision, uncertainty, partial truth, and approximation.  Well suited for real world problems where ideal models are not available.

ELEMENTS OF SOFT COMPUTING

FUZZY LOGIC SYSTEMS

ARTIFICIAL NEURAL NETWORKS

EVOLUTIONARY ALGORITHMS

Fundamentals of Neural Networks What is Neural Network ?  An information processing model that is inspired by the way

biological nervous system such as the brain, process information.  A neural network is an artificial representation of the human brain that tries to simulate its learning process.  An artificial neural network (ANN) is often called a "Neural

Network“ or simply Neural Net (NN).

Fundamentals of Neural Networks  ANN is an interconnected group of artificial neurons that

uses a mathematical model for information processing.  An ANN is configured for a specific application through a

learning process.

Why Neural Network?  Neural

Networks

follow

a

different

paradigm

for

computing.  The conventional computers are good for - fast arithmetic Not so good for - interacting with noisy data or data from

the environment  The neural network systems help where we cannot formulate an algorithmic solution

Advantages Adaptive learning – Ability to learn Self organization – Creates its own Real time operation

Applications Medical diagnosis Recognition of Photos and fingerprints Speech recognition Load Forecasting Weather Forecasting

Appraisal and valuation of property Machinery control etc.,

Biological Neuron Terminal Branches of Axon

Dendrites Synapse Nucleus

Axon

Artificial Neuron Terminal Branches of Axon

Dendrites x1 w1 x2 x3

w2 w3

S Axon

wn xn

ANN - HISTORY

….

ANN – HISTORY

BASIC MODELS OF ANN The model's synaptic interconnections

The training or learning rules adopted for updating and adjusting the connection weights

Their activation functions.

CONNECTIONS  Single-layer feed-forward network  Multilayer feed-forward network  Single node with its own  Single-layer recurrent network  Multilayer recurrent network

Single-layer feed-forward network

Multilayer feed-forward network

Single node with its own

Single-layer recurrent network

Multilayer recurrent network

LEARNING  Supervised Learning  Unsupervised Learning  Reinforcement Learning

Supervised Learning

Unsupervised Learning

Reinforcement Learning

ACTIVATION FUNCTIONS Identity function:

Binary Step function:

ACTIVATION FUNCTIONS….

Bipolar Step function:

Sigmoidal functions: Binary Sigmoid function:

ACTIVATION FUNCTIONS….

Bipolar Sigmoid function:

Ramp function:

TERMINOLOGIES IN ANN  Weights  Bias  Threshold  Learning Rate  Momentum Factor  Vigilance Factor

PERCEPTRON  Perceptron solution [Rosenblatt, 1958]

 Weights and thresholds can be determined analytically or by a learning algorithm

 Continuous, bipolar and multiple-valued versions